[Kamailio-Users] ERROR:mi_fifo:mi_fifo_server: fifo command must begin with :: early_dialogs (1.5 rev 5834)

[Iñaki Baz Castillo]

2009/10/28 Daniel-Constantin Mierla <miconda at gmail.com>:
>> The error remains and it just affects to dialog module. I log in a file
>> when the Munin plugin performing the fifo command gets a dialg_xxxx value
>> greater than 200 (note that there is no traffic in this Kamailio !!!), and
>> this is what I've got from past friday (3 days ago) having the latest 1.5
>> version installed:
>>
>> Sat Oct 17 00:50:03 CEST 2009: dialogs_early = 134217728
>> Sat Oct 17 00:50:03 CEST 2009: dialogs_total = 134217728
>> Sat Oct 17 16:50:03 CEST 2009: dialogs_early = 2820392
>> Sat Oct 17 16:50:03 CEST 2009: dialogs_total = 2820392
>>
>>
> This is interesting as you say there is no traffic on server. Since is a
> testing system I guess we can play a bit with it.

Well, no it's already on production, anyhow I can experiment a bit on it :)



> The error might be due to
> a bug in FIFO command parsing or a bug in printing the command to fifo file.

If it helps, I could reproduce the problem without the munin plugin,
just by running the command "kamctl fifo get_statistics
active_dialogs" several times. Ramdomly it returns an error "fifo
command must begin with :: total_size".



> The best is to add a debug message that will print the entire fifo command
> received by kamailio. Should I do a patch for you?

I will try by myself, thanks.


> More strange is with dialog value as there is no traffic, that value should
> not be affected at all and should be zero. If you do kamctl get_statistics
> all from command line, do you get same value?

The main problem is that this issue occurs jsut *some* times. Note
that munin plugins are executed each 5 minutos and the MI command just
fails "some times", so it would be really impossible to check if
"kamctl get_statistics" also fails.

However, about the big dialogs numbers the command returns, I extract
that value as follows (note that I did some modifications in the munin
plugin trying top detect the isssue):

-------------------
dialogs_early=$(kamctl fifo get_statistics early_dialogs) 2>>
/tmp/munin-kamailio.error
retcode=$?
if [ $retcode -ne 0 ] ; then
  echo "$(date): kamctl fifo get_statistics early_dialogs -> retcode =
$retcode"  >> /tmp/munin-kamailio.error
fi

# This is not a server in production so dialog_early must be near 0:
if [ $dialogs_early -ge 200 ] ; then
  echo "$(date): dialogs_early = $dialogs_early" 2>&1 >>
/tmp/munin-kamailio.error
fi
------------------


And the output I get in  /tmp/munin-kamailio.error is

Sat Oct 17 00:50:03 CEST 2009: dialogs_early = 134217728
Sat Oct 17 16:50:03 CEST 2009: dialogs_early = 2820392
Fri Oct 23 12:45:04 CEST 2009: dialogs_early = 2863080


This is:

The command "dialogs_early=$(kamctl fifo get_statistics early_dialogs)
2>> /tmp/munin-kamailio.error" ALWAYS returns 0 (correct return code
in bash) so it never "fails".

The value the above command returns is obviously wrong (134217728, 2820392...).




-- 
Iñaki Baz Castillo
<ibc at aliax.net>